Paclitaxel chemotherapy disrupts behavioral and molecular circadian clocks in mice

Abstract

Cancer patients experience circadian rhythm disruptions in activity cycles and cortisol release that correlate with poor quality of life and decreased long-term survival rates. However, the extent to which chemotherapy contributes to altered circadian rhythms is poorly understood. In the present study, we examined the extent to which paclitaxel, a common chemotherapy drug, altered entrained and free-running circadian rhythms in wheel running behavior, circulating corticosterone, and circadian clock gene expression in the brain and adrenal glands of tumor-free mice. Paclitaxel injections delayed voluntary wheel running activity onset in a light-dark cycle (LD) and lengthened the free-running period of locomotion in constant darkness (DD), indicating an effect on inherent suprachiasmatic nucleus (SCN) pacemaker activity. Paclitaxel attenuated clock gene rhythms in multiple brain regions in LD and DD. Furthermore, paclitaxel disrupted circulating corticosterone rhythms in DD by elevating its levels across a 24-hour cycle, which correlated with blunted amplitudes of Arntl, Nr1d1, Per1, and Star rhythms in the adrenal glands. Paclitaxel also shortened SCN slice rhythms, increased the amplitude of adrenal gland oscillations in PER2::luciferase cultures, and increased the concentration of pro-inflammatory cytokines and chemokines released from the SCN. These findings indicate that paclitaxel disrupts clock genes and behavior driven by the SCN, other brain regions, and adrenal glands, which were associated with chemotherapy-induced inflammation. Together, this preclinical work demonstrates that chemotherapy disrupts both central and peripheral circadian rhythms and supports the possibility that targeted circadian realignment therapies may be a novel and non-invasive way to improve patient outcomes after chemotherapy.

1. Introduction

Circadian rhythms are present in nearly every mammalian cell and orchestrate cellular, hormonal, and behavioral processes to anticipate changes to day length, synchronizing the body to the 24-h period of the Earth’s rotation. These oscillations are driven by an established transcriptional-translational feedback loop (Jin et al., 1999; Kume et al., 1999). Briefly, the genes Arntl (Bmal1) and Clock are transcribed and translated into the proteins BMAL1 and CLOCK. BMAL1 and CLOCK subsequently heterodimerize, translocate to the nucleus, and bind to E-box transcription factor binding sites in the promoters of clock-controlled genes (Gekakis et al., 1998; Jin et al., 1999). The core clock genes Period (Per1/Per2/Per3) and Cryptochrome (Cry1/Cry2) are transcribed by CLOCK/BMAL1 dimers, translated into protein, and then form PER/CRY dimers (Jin et al., 1999; Kume et al., 1999). PER/CRY heterodimers then translocate back to the nucleus where they inhibit CLOCK/BMAL1 transcriptional activity and thus their own transcription (Jin et al., 1999; Kume et al., 1999); this negative transcriptional-translational feedback loop occurs with an approximate 24-h period. The amplitude of Bmal1, and therefore clock-controlled transcriptional rhythms, is further modulated by a secondary transcriptional feedback loop that activates (RORα, coded by Rora, Akashi and Takumi, 2005) or inhibits (REV-ERBα, coded by Nr1d1, Preitner et al., 2002) Bmal1 transcription.

The suprachiasmatic nucleus (SCN) of the hypothalamus in the brain acts as a “master clock” by eliciting endocrine and autonomic signals to synchronize (i.e., “entrain”) locomotor, hormonal, and cellular rhythms throughout various tissues in the body (Balsalobre et al., 2000, reviewed in Dibner et al., 2010). The retina then sets the phasing of the inherent pacemaker activity of the SCN by a direct synaptic projection to set timing throughout the body in accordance with available environmental light cues. Notably, these oscillations are endogenous as they persist in constant darkness with a period close to 24 h, as evidenced by voluntary wheel running rhythms, the behavioral output of the SCN (Bunger et al., 2000).

Clock gene rhythms have been characterized in brain regions outside of the SCN, including the hippocampus (Chun et al., 2015; Jilg et al., 2010) and cerebral cortex (Wakamatsu et al., 2001; Yang et al., 2007), where these molecular rhythms are coupled to time-of-day differences in other behaviors, including cognitive function (Bering et al., 2017; Snider et al., 2016). Adrenal gland function in the periphery is also tied to its circadian clock. The gene encoding steroidogenic acute regulatory protein (StAR, encoded by Star), responsible for the synthesis of the primary glucocorticoid (corticosterone; CORT) in mice, is rhythmically expressed by CLOCK/BMAL1 transcription and correlates with robust circadian rhythms of CORT release (Son et al., 2008). Moreover, intrinsic CORT rhythms persist independently of the master clock (Szafarczyk et al., 1979). Thus, the function of circadian clocks throughout the body is critical for normal physiological function.

Chemotherapeutics are instrumental pharmacological cancer treatments that disable cell division in highly proliferative cells (e.g., tumor cells). However, chemotherapy also causes cell death of healthy, proliferative cells in the body, which ultimately contributes to debilitating side effects, including fatigue (Li et al., 2018; Savard et al., 2009; Schmidt et al., 2016), cognitive deficits (Cheung et al., 2015; Kesler et al., 2013), and decreased overall quality of life in cancer patients (Sultan et al., 2017). Furthermore, flattened glucocorticoid and activity circadian rhythms have been reported in chemotherapy-treated cancer patients and survivors (Abercrombie et al., 2004; Cuneo et al., 2017; Schmidt et al., 2016; Touitou et al., 1996) which are associated with inflammation (Kober et al., 2016) and reduced long-term patient survival (Mormont et al., 2000; Sephton et al., 2000). Thus, as poor survival rates are correlated with dysregulated circadian rhythmicity, it is important to understand the origins of circadian disruption in cancer patients. Moreover, as chemotherapy has been demonstrated to induce signaling of the pro-inflammatory cytokines TNF-α and IL-1β (Smith et al., 2014) and TNF-α can interfere with clock gene expression (Cavadini et al., 2007), it is possible that chemotherapy-induced inflammation may contribute to circadian misalignment in cancer patients.

However, studies investigating the extent to which chemotherapy affects circadian rhythms are scarce. In tumor-naïve mice, there are few studies reporting chemotherapeutics inducing changes in circadian rhythms (reviewed in Smith et al., 2014), including vinorelbine- and gemcitabine-induced changes to body temperature and activity rhythms (Li and Levi, 2007; Li et al., 2002). However, to our knowledge no studies have examined the extent to which chemotherapy affects endogenous rhythms in master clock-driven behavior in constant darkness. Other chemotherapeutics have been reported to induce decreased locomotor activity in a light-dark cycle and dysregulated SCN clock gene rhythms (5-fluoruracil, Terazono et al., 2008), circulating corticosterone rhythms (irinotecan, Ahowesso et al., 2011), and circadian gene expression in peripheral tissues (cisplatin, Cao et al., 2018). However, circadian disruption caused by paclitaxel, a common taxane chemotherapy classified as an essential medicine by the World Health Organization for treating breast, cervical, non-small cell lung, and ovarian cancers (Organization, 2019), has not been evaluated. Moreover, chemotherapy effects have not been studied in multiple tissues simultaneously to understand the distribution of cancer treatment on master and auxiliary circadian clocks. Here, we assessed the extent to which paclitaxel disrupts voluntary wheel running rhythms in a light-dark cycle and in constant darkness to understand if this chemotherapeutic can affect SCN-driven behavior in female mice, as paclitaxel is prescribed for cancers which occur predominantly in women. We also examined the extent to which paclitaxel disrupts circadian oscillations in CORT release and clock gene expression in the brain and adrenal glands. Finally, the direct effects of paclitaxel on the rhythmicity of SCN and adrenal gland cultures were tested in vitro, and cytokines and chemokines were examined in SCN culture media to examine the effects of chemotherapy-induced inflammation on circadian timing.

2. Materials and Methods

2.1. Mice

Seven to nine-week-old female C57BL/6 mice (Charles River, Wilmington, MA, USA, RRID: IMSR_CRL:027) were acclimated to a temperature-controlled (22 ± 1 °C) vivarium for 1 week. Two to six-month-old, female PER2::luciferase mice (Jackson Laboratories, Bar Harbor, ME, USA, RRID: IMSR_JAX:006852) (Yoo et al., 2004) were used for in vitro PER2::luciferase rhythmicity studies. Standard rodent chow (Harlan 7912) and water were available ad libitum throughout all experiments. All experimental procedures were performed with prior approval from the Ohio State University Institutional Animal Care and Use Committee and based on standards listed in the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NRC et al., 2011). Efforts were made to reduce individual animal suffering and the number of animals whenever possible. Mice were singly housed in running wheel experiments and housed with two mice per cage in all other experiments.

2.2. Experimental design.

2.2.1. Voluntary wheel running rhythms in LD.

Mice were given 14 d to acclimate to running wheels. Following acclimation, in-cage voluntary wheel running was recorded for three baseline days in a 14 h-light: 10 h-dark (14:10 LD) cycle and for the remainder of the experiment. Next, mice were randomized into paclitaxel chemotherapy or vehicle treatment groups (n = 8/group) and were injected over an 11-d period as described below. Voluntary wheel running was recorded during the entire treatment period and for a 20-d recovery period (Fig. 1A). Recording was briefly disrupted for a 2-d period during the recovery period, which did not significantly affect analyses of total activity or rhythmicity.

Figure 1. Running wheel activity onset is decreased by paclitaxel chemotherapy.

A. Timeline of chemotherapy effects on LD running wheel rhythms experiment. B. Median, double-plotted actograms averaged from vehicle- and chemotherapy-treated mice. Red dots indicate times of each injection. C. Total running wheel activity was changed by an interaction with time and chemotherapy in a light-dark cycle. Arrows indicate days of vehicle or paclitaxel injections. D. The onset of wheel running activity, as measured by phase angle of entrainment, was delayed in paclitaxel-treated mice. N = 8 mice/group, *p < 0.05.

2.2.2. Circadian gene expression, CORT, and cytokine rhythms in LD.

Mice were randomized into groups and injected with paclitaxel chemotherapy (n = 35) or vehicle (n = 48) in a 12 h-light: 12h-dark (12:12 LD) cycle. Twenty-four hours after the final injection, mice were randomized into six subgroups (n =5-6/group/time point) and sacrificed at 4 h-intervals over a 24-h period beginning at 2 h after lights on (i.e., zeitgeber time [ZT] 2): ZT 2, 6, 10, 14, 18, and 22 (Fig. 2A) to analyze rhythms in circadian gene expression, circulating corticosterone (CORT), and circulating cytokines/chemokines.

Figure 2. Paclitaxel effects on clock gene expression in the brain in a light-dark cycle.

A. Experimental timeline of chemotherapy effects on circulating corticosterone and gene expression rhythms in a light-dark (LD) cycle. B. Hypothalamic Per1 was rhythmic in control but not chemotherapy-treated mice. Chemotherapy lowered the amplitude of hypothalamic Per2 rhythms by decreased night expression. C. Hippocampal Arntl rhythms were altered by chemotherapy. Hippocampal Nr1d1 and Per2 rhythms were present in both vehicle- and chemotherapy-treated mice, with a significant decrease in Per2 expression at ZT 2. D. Frontal cortex Nr1d1, Per1, and Per2 rhythm amplitudes were flattened by chemotherapy while Arntl rhythms were consistent between groups. Solid lines indicate statistically significant rhythms, dashed lines indicate non-rhythmic transcripts. N =5-6/group/time point, *p < 0.05.

2.2.3. Voluntary wheel running rhythms in DD.

All mice (n = 24) were acclimated to in-cage running wheels and then recorded for 7 d in a 12:12 LD cycle (baseline). Next, mice were placed in constant darkness (DD) and wheel running activity was recorded for the remainder of the experiment. After one initial week of recording baseline DD activity, mice were then randomized into groups and injected with either paclitaxel chemotherapy or vehicle for controls (n = 12/group) over an 11-d period. Injections were made at inconsistent times so as not to alter free-running rhythms by treatment timing. Following treatment, wheel running activity was recorded for a 17-d recovery period in DD (Fig. 4A).

Figure 4. Paclitaxel lengthens the free-running period of voluntary wheel running during recovery.

A. Experimental timeline of constant darkness running wheel rhythmicity experiment. B. Median, double-plotted actograms of averaged activity from vehicle- and chemotherapy-treated mice. Red dots indicate times of each injection. C. Total wheel running revolutions were decreased during the time of injections (days indicated by arrows) and on the week following injections. D. Chemotherapy lengthened free-running periods of running wheel rhythms during the second week of recovery. N = 12/group, *p < 0.05.

2.2.4. Circadian gene expression, CORT, and cytokine rhythms in LD.

Mice (n = 74) were placed in constant darkness (DD) for 1 week. Paclitaxel chemotherapy or vehicle were then injected as described below with injections made at inconsistent times in order to avoid influencing free-running rhythms. One day after the final treatment, mice were sacrificed at 4-h intervals (n = 6-8/time point/group) beginning at circadian time (CT) 2, defined as 2 h after the time in which lights were turned on during a LD cycle (i.e., “subjective day”) and continuing until CT 22 (Fig. 5A). Rhythms in circadian gene expression, circulating corticosterone (CORT), and circulating cytokines/chemokines were analyzed.

Figure 5. Paclitaxel chemotherapy effects on clock gene expression in the brain in constant darkness.

A. Experimental timeline of chemotherapy effects on circulating corticosterone and gene expression rhythms in a light-dark (LD) cycle. B. Hypothalamic clock gene expression was increased at CT14 in chemotherapy-treated mice. Chemotherapy also increased the hypothalamic Per2 rhythm amplitude. C. Chemotherapy abolished hippocampal Nr1d1 rhythms and altered Per1 expression in the subjective night. D. Chemotherapy altered expression of multiple frontal cortex clock genes in the early subjective day and blunted the amplitudes of Arntl, Nr1d1, and Per2 rhythms. Solid lines indicate statistically significant rhythms, dashed lines indicate non-rhythmic transcripts. N = 6-8/time point/group, *p < 0.05; **p < 0.01; ****p < 0.0001.

2.2.5. PER2::LUC adrenal gland and SCN slice rhythms.

Nine female PER2::luciferase (PER2::LUC) were placed in constant darkness (DD) for two days (Fig. 7A). Under constant darkness, paclitaxel or vehicle injections were injected as described below with injections given at inconsistent times between injections to avoid influencing free-running rhythms. Mice were sacrificed 20-21 h after the final injection and both adrenal glands and suprachiasmatic nucleus (SCN) slices were cultured on inserts containing culture media and beetle luciferin. A subset of SCN slices and single adrenal glands from mice treated with vehicle in vivo were supplemented with 10 μM paclitaxel to test the direct effects of chemotherapy in vitro. Cultured sections were placed in a LumiCycle luminometer (Actimetrics, Wilmette, IL, USA) inside a CO₂ incubator and PER2::LUC rhythms were analyzed for 4 d.

Figure 7. Paclitaxel shortens SCN period length and increases adrenal gland rhythm amplitude in vitro.

A. Timeline of PER2::LUC experiment. B. Paclitaxel shortened the period of SCN circadian rhythms but did not alter the phase or amplitude of rhythms. C. Chemotherapy increased the amplitude of adrenal PER2::LUC rhythms but did not alter period or phase of rhythms. D. Chemotherapy administered in vivo increased cytokines and chemokines in culture media from SCN slices, which was not increased by chemotherapy administered in vitro. N = 3-5/group, *p < 0.05; **p < 0.01; ***p < 0.001.

2.3. Chemotherapy treatment.

Paclitaxel chemotherapy (Millipore-Sigma, Burlington, MA, USA, cat #: T7191) was dissolved in Cremophor EL:PBS solution (Millipore-Sigma, cat #: 238470-1SET) as previously described (Loman et al., 2019). Briefly, chemotherapy (30 mg/kg in 100 μl, i.p.) or vehicle was injected every other day for a total of six doses. Notably, the dose and number of chemotherapy cycles in human can vary depending on severity of cancer progression as well as response to therapy; thus, many paclitaxel doses and regimens have been used in a clinical setting. We therefore determined our dose strength and number of doses in consultation with a medical oncologist and subsequently used the animal equivalent dosage of a 90 mg/m² weekly paclitaxel dose in humans, given the clinical range of 80-100 mg/m² (Schott, 2021). Using a previously validated conversion rate (Nair and Jacob, 2016), we calculated the following mouse equivalent dose:

$$\begin{gathered} 90{\text{mg/m}}^2(\text{human dose})∕37{\text{human K}}_{\mathrm{m}}=2.43\text{mg/kg human dose} \\ \text{mouse dose (mg/kg)}=2.43\text{mg/kg}∕(3{\mathrm{K}}_{\mathrm{m}}\text{mouse}∕37{\mathrm{K}}_{\mathrm{m}}\text{human})=30\text{mg/kg} \end{gathered}$$

Women with breast cancer commonly receive 4-8 cycles of adjuvant chemotherapy (including paclitaxel); thus, we chose 6 “cycles” of paclitaxel as a midpoint for a mouse equivalent of paclitaxel. Given the difference in mouse to human lifespan, we scaled our timing of dosage to every other day compared to weekly accordingly (2 mouse months is approximately 10 human years). Mice were pseudorandomized into chemotherapy or vehicle groups based on equivalent initial body mass. Body mass was measured every other day; one mouse across all experiments lost > 10 % of body mass over a 2-d period and was excluded from further analyses.

2.4. Voluntary wheel running rhythms.

Prior to recording, mice were acclimated to in-cage running wheels (Starr Life Sciences, Oakmont, PA, USA) for 3 d. Continuous, 24-h recordings of voluntary wheel running revolutions were made every 5 min using a probe attached to the running wheel and VitalView 5.1 software (Starr Life Sciences, RRID: SCR_014497). Activity plots (actograms), activity onsets, and total 24-h wheel revolutions were analyzed during baseline, treatment, and recovery periods using ClockLab Analysis 6.1 software (Actimetrics, RRID: SCR_014309). The period of voluntary wheel running rhythms was analyzed using a chi-square periodogram in ClockLab Analysis 6.1 and total 24-h wheel revolutions were calculated for each day analyzed. The phase angle of entrainment (LD cycles only) was calculated as the difference (in hours) between the time when lights turned off and the onset of running wheel activity.

2.5. Tissue/plasma collection.

Mice were euthanized using rapid decapitation under either white (ZT 2/6/10 time points in Figs. 2A & 3A) or dim, dark room-grade red light (all other time points in Figs. 2, 3, 5, & 6). Trunk blood was immediately collected through heparinized Natleson capillary tubes (Kimble Chase, Rockwood, TN, cat #: 14705-018) into microcentrifuge tubes kept on ice. Next, a ~0.5 cm³ cube of whole hypothalamus was freshly dissected from the brain along with both hippocampi and a ~0.2 cm³ portion of frontal cortex from both hemispheres. These brain regions were chosen due to the importance of the hippocampus and frontal cortex (memory and cognitive processes) and the hypothalamus (the region containing the SCN). Both adrenal glands were then dissected and weighed, and all tissues were flash frozen on dry ice followed by storage at −80 °C until RNA extraction. Following each tissue collection, whole blood was centrifuged (595 x g, 20 min, 4 °C) to obtain plasma, which was stored at −80 °C.

Figure 3. Paclitaxel effects on circulating corticosterone and adrenal gland clock gene expression in a light-dark cycle.

A. Corticosterone rhythms had increased amplitude following chemotherapy treatment. B. Nr1d1, Per2, and Star rhythm amplitudes were flattened by paclitaxel treatment. Solid lines indicate statistically significant rhythms, dashed lines indicate non-rhythmic transcripts. N =5-6/group/time point, *p < 0.05.

Figure 6. Paclitaxel chemotherapy effects on circulating corticosterone and adrenal gland clock gene expression in constant darkness.

A. Chemotherapy ablates circulating CORT rhythms by elevating CORT concentrations at multiple times of day. B. Chemotherapy decreased rhythm amplitudes of Arntl, Nr1d1, Per1, and Star in the adrenal gland while also altering clock gene expression at various times of day. Solid lines indicate statistically significant rhythms, dashed lines indicate non-rhythmic transcripts. N = 6-8/time point/group, *p < 0.05; **p < 0.01; ***p < 0.001.

2.6. Reverse transcription quantitative PCR (RT-qPCR).

RNA was extracted from adrenal glands, frontal cortex, hippocampus, and hypothalamus using RNeasy Mini kits (QIAGEN, Hilden, Germany, cat #: 74106). RNA concentrations were measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA, cat #: ND-ONE-W) with 260/280 ratios ~1.8-2.0. One μg of cDNA per sample was reverse transcribed (RT) from extracted RNA using qScript cDNA Synthesis kits (QuantaBio, Beverly, MA, USA, cat #: 95048-500). Gene expression was determined using quantitative polymerase chain reaction (qPCR) reactions with a QuantStudio 5 384-well thermal cycler (Thermo Fisher Scientific, cat #: A28140) and the following TaqMan probes (Thermo Fisher Scientific): Arntl (cat #: 4331182, assay: Mm00500226_m1), Gapdh (cat #: 4331182, assay: Mm99999915_g1), Hprt (cat #: 4331182, assay: Mm03024075_m1), Nr1d1 (cat #: 4331182, assay: Mm00520708_m1), Per1 (cat #: 4331182, assay: Mm00501813_m1), Per2 (cat #: 4331182, assay: Mm00478099_m1), and Star (cat #: 4331182, assay: Mm00441558_m1). Relative gene expression was ultimately calculated using the comparative C_T method (2^−ΔΔCT) by subtracting C_T values of circadian genes of interest (Arntl, Nr1d1, Per1, Per2, and Star) to the geometric mean of the housekeeping genes glyceraldehyde 3-phosphate dehydrogenase (Gapdh) and hypoxanthine-guanine phosphoribosyltransferase (Hprt), and then normalizing these ΔC_T values to the average ΔC_T of vehicle-treated mice sacrificed at ZT 2 or CT 2 to determine fold change in gene expression (2^−ΔΔCT).

2.7. Plasma CORT and cytokine/chemokine concentrations.

Corticosterone (CORT) was measured in duplicate in all samples via EIA according to the manufacturer’s instructions (Enzo Life Sciences, Farmingdale, NY, USA, cat #: ADI-900-097). Samples were run using a 1:40 dilution with intraassay and interassay variations <10%. Circulating cytokines and chemokines (IL-1β, IL-2, IL-4, IL-5, IL-6, IL-10, IL-12p70, IFN-γ, TNF-α, CXCL1 [KC/GRO], CXCL2 [MIP-2α], CXCL10 [IP-10], CCL2 [MCP-1 α], CCL3 [MIP-1]) were measured using a U-PLEX assay (Meso Scale Discovery, Rockville, MD, USA, cat #: K15069L-1) according to the manufacturer’s instructions on a QuickPlex SQ 120 instrument (Meso Scale Discovery, cat #: AI0AA-0, RRID: SCR_020304). Different analytes were measured between experiments based on kit availability at the time the experiment was conducted. All samples were run in duplicate with interassay variations < 10%.

2.8. PER2::luciferase adrenal gland and SCN slice cultures.

Twenty to 21 h following the final paclitaxel (n = 5) or vehicle (n = 4) injection, PER2::luciferase (PER2::LUC) mice were sacrificed under dim red light via rapid decapitation. Whole brains and both adrenal glands were extracted and placed into ice cold, sterile, oxygenated artificial cerebrospinal fluid (aCSF) at pH 7.40 with the following solutes: 120 mM NaCl, 3.5 mM KCl, 10 mM HEPES, 1.2 mM NaH₂PO₄ monohydrate, 0.5 mM CaCl₂ dihydrate, 3 mM MgSO₄ heptahydrate, 32.3 mM NaHCO₃, 10 mM dextrose. Using a vibratome (Leica Microsystems, Wetzlar, Germany, cat #: 14048142065), 180 μM-thick coronal sections containing the SCN were obtained. SCN slices and whole adrenal glands were submerged in dissection media (aCSF, 1% penicillin/streptomycin [Thermo Fisher Scientific, cat #: 15140122], 0.06% nystatin [Millipore-Sigma, cat #: N1638], 0.1 μM MK 801 [Tocris, Bristol, UK, cat #: 0924]), and bisected adrenal glands and ~ 0.5 mm x 0.5 mm x 180 μM hypothalamic squares containing the SCN were obtained. SCN sections were obtained using microdissection scissors under a dissecting microscrope using the optic chiasm, the outline of the suprachiasmatic nuclei, and the third ventricle as landmarks. Individual SCN squares or bisected adrenals were then placed on cell culture inserts (Millipore-Sigma, cat #: PICM0RG50, RRID: SCR_015799) in 35 mm dishes (Falcon, Corning, NY, cat #: 351008) containing 1 mL culture media (SCN culture media: Neurobasal A minus phenol red [Thermo Fisher Scientific, cat #: 12349015], 5% Horse Serum [Millipore-Sigma, cat #: H1270], 2% B27 Supplement [Thermo Fisher Scientific, cat #: 17504044], 1% penicillin/streptomycin, 0.5 mM GlutaMAX [Thermo Fisher Scientific, cat #: 35050061], 0.06% nystatin, 0.1 μM MK 801, 0.1 mM beetle luciferin [Promega, Madison, WI, USA, cat #: E1602]; adrenal gland culture media: DMEM with high glucose minus phenol red [Thermo Fisher Scientific, cat #: 21063029], 1% penicillin/streptomycin, 0.06% nystatin, 0.1 mM beetle luciferin). In a subset of SCN and adrenal gland cultures processed from mice treated with vehicle in vivo, 10 μM of paclitaxel was added to the culture media (vehicle + chemo in vitro group), as this concentration has been shown to be pharmacologically active in vitro (Schiff and Horwitz, 1980). Dishes were sealed with vacuum grease and sterile 40 mm circular coverslips to prevent media evaporation in a laminar flow biosafety cabinet. All dishes were then placed in a light-tight LumiCycle luminometer (Actimetrics) in an cell culture incubator containing 5% CO₂ at 33°C. Cultures were given 24 h to recover, and then PER2::luciferase rhythms were recorded at 10-min intervals for 4 d using LumiCycle Analysis 2 (Actimetrics). Due to poor seals on two dishes, two chemo in vivo slices were discarded resulting in these final sample sizes: SCN vehicle: n = 4; SCN vehicle + chemo in vitro: n = 4; SCN chemo in vivo: n = 5; adrenal gland vehicle: n = 4, adrenal gland vehicle + chemo in vitro: n = 4; adrenal gland chemo in vivo: n = 3.

2.9. Rhythmicity analysis.

Statistically significant rhythms were quantified using the R package MetaCycle (Wu et al., 2016). MetaCycle parameters were as follows: p < 0.05, minper = 20, maxper = 26, cycMethod= c (“ARS,” “JTK,” “LS”), analysisStrategy = “auto,” outputFile = TRUE, outIntegration = “both,” adjustPhase = “predictedPer,” combinePvalue = “bonferroni”, weightedPerPha = FALSE, ARSmle = “auto,”, and ARSdefaultPer = 24. Period, phase, and amplitude of rhythms were recorded along with a p-value to indicate a statistically significant rhythm. Using these parameters, the presence of near 24-h rhythms in circulating concentrations of corticosterone, chemokines/cytokines, and circadian gene expression were assessed in vehicle- and chemotherapy-treated mice.

PER2::luciferase rhythms were analyzed by using LumiCycle Analysis 2 (Actimetrics) to generate detrended data using a running average of 24 h and smoothing 25 data points. The detrended data were then exported to MetaCycle for analysis using the above parameters with the following adjustments: minper = 12, maxper = 48.

2.10. Statistical analysis.

Statistical comparisons of behavioral rhythm parameters, gene expression, and circulating corticosterone were analyzed using Student’s t-test using GraphPad Prism version 8.2.1 (GraphPad Software, San Diego, USA) when variance was normal, or Mann-Whitney U tests were used when variance was non-parametric. For total wheel running activity analysis, a repeated measures two-way ANOVA was performed using GraphPad Prism. Outliers were removed from datasets using Grubb’s test with α = 0.05. For gene expression, circulating cytokine/chemokine, and corticosterone rhythmicity analyses, Student’s t-test was used for pairwise comparisons at each time point between vehicle and chemotherapy groups to assess differences between rhythmic patterns. One-way ANOVA followed by Bonferroni’s multiple comparisons test was used for distinguishing statistically significant differences in PER2::LUC rhythms and circulating cytokines/chemokines secreted in the surrounding culture media. Data were determined to be statistically significant at p ≤ 0.05 and all graphs are presented as mean ± standard error of the mean using GraphPad Prism.

3. Results

3.1. Chemotherapy decreases running wheel activity and delays activity onset in a light-dark cycle.

Following acclimation to in-cage running wheels, baseline voluntary wheel running activity was recorded for 3 d (Fig. 1A-B). As expected, all mice began wheel running activity closely to the time of lights off during the baseline period (Fig. 1B). As predicted based on our previous studies (Loman et al., 2019; Sullivan et al., 2020), chemotherapy induced fatigue in this light-dark cycle as measured by decreased voluntary wheel running during treatment (interaction between day and drug: F_{30, 360} = 1.963, p < 0.005, Fig. 1C). Moreover, while the onset and period of wheel running activity did not differ between groups during the baseline or weeks of recovery (p > 0.05), the onset of activity as measured by phase angle of entrainment was delayed during the time of paclitaxel injections (t₁₄ = 2.427, p < 0.05, Fig. 1D).

3.2. Chemotherapy dampens the amplitude of brain clock gene rhythms in a light-dark cycle.

Clock gene expression was examined in the brain following chemotherapy treatment in a light-dark cycle (Fig. 2A). Hypothalamic circadian rhythms of Arntl (BMAL1) and Nr1d1 (REV-ERBα) were not detected in either group (Fig. 2B). However, paclitaxel ablated hypothalamic Per1 rhythms and decreased the rhythm amplitude of hypothalamic Per2 due to decreased expression in the night at zeitgeber time (ZT) 18 (t₈ = 2.65, p < 0.05, Fig. 2B, Supplementary Table 1). In the hippocampus, paclitaxel altered Arntl rhythms, while Nr1d1 and Per2 rhythms were detected in both vehicle- and chemotherapy-treated mice (p < 0.05, Fig. 2C, Supplementary Table 1). Hippocampal Per2 expression was decreased by chemotherapy in the early light phase at ZT 2 (t₉ = 2.94, p < 0.05), but chemotherapy did not alter hippocampal Per1 gene expression, nor did Per1 expression meet criteria for circadian rhythmicity in either group (p > 0.05, Fig. 2C). In the frontal cortex, Arntl, Nr1d1, Per1, and Per2 rhythms were all rhythmic and this rhythmicity was not affected by chemotherapy (p < 0.05 all genes). However, the amplitude of Nr1d1, Per1, and Per2 rhythms were each blunted by paclitaxel (Fig. 2D, Supplementary Table 1). These amplitude effects were driven by differences between groups in frontal cortex Nr1d1 (ZT 6: t₁₂ = 2.57, p < 0.05; ZT14: t₈ = 2.68, p < 0.05) and Per2 (ZT 18: t₁₂ = 2.45, p < 0.05; Fig. 2D).

3.3. Chemotherapy dampens the amplitude of adrenal gland clock gene rhythms and increases the amplitude of corticosterone rhythms in a light-dark cycle.

Circulating corticosterone (CORT) was rhythmic in both vehicle- and chemotherapy treated mice (p < 0.05). However, paclitaxel increased the amplitude of circulating corticosterone (CORT) rhythms in a light-dark cycle (Fig. 3A, Supplementary Table 1). This increased CORT amplitude in chemotherapy-treated mice was driven by a trend of increased circulating CORT at ZT 10 (p = 0.054) and at ZT18 (p = 0.051; Fig. 3A). Circulating cytokine and chemokine (IL-1β, IL-6, IL-10, TNF-α, CXCL1, CXCL2, CXCL10, CCL2, CCL3) rhythms were not detected in either vehicle or paclitaxel-treated groups for all analytes, with the exception of a ~24h rhythm detected in the vehicle group for TNF-α (p < 0.05, Supplementary Fig 1).

Adrenal gland clock gene rhythms were detected in chemotherapy-treated mice (p < 0.05), but the amplitude of Nr1d1, Per2, and Star rhythms were reduced by paclitaxel (Fig. 3B, Supplementary Table 1). While there were no statistically significant differences in Star expression between treatment groups at individual time points, reduced Nr1d1 and Per2 rhythm amplitudes were driven by differences at the following time points: Nr1d1 ZT 18: t₁₂ = 2.45, p < 0.05; Per2 ZT 10: t₈ = 2.52, p < 0.05, ZT 14: t₁₀ = 2.75, p < 0.05 (Fig. 3B).

3.4. Chemotherapy lengthens running wheel rhythmicity in constant darkness.

Following habituation to running wheels, baseline running wheel activity was recorded for one week in a 12:12 LD cycle followed by one week in constant darkness (DD; Fig. 4A-B). Total activity remained consistent between experimental groups prior to treatment, and the period of wheel running rhythms was 24 h in both groups during the LD cycle period and did not differ between groups (23.7-23.8 h) during the baseline DD period prior to treatment (p > 0.05, Fig. 4B-D). At the time of injections and continuing through the first week of recovery, chemotherapy lowered the total amount of running wheel activity (interaction: time x drug: F_{42, 924} = 7.00, p < 0.0001) but did not alter free-running wheel rhythms (p = 0.1, Fig. 4B-D). However, as total activity returned to control levels during the second week of recovery (i.e., Week 6), chemotherapy lengthened the free-running period of wheel running activity (U = 32.5, n₁ = 11, n₂ = 12, p < 0.05, Fig. 4B, 4D).

3.5. Chemotherapy attenuates rhythms in brain clock gene expression in constant darkness.

After one week in constant darkness, mice were injected with chemotherapy or vehicle and were then sacrificed at 4-h intervals on the day following the final injection to analyze clock gene rhythms throughout the brain (Fig. 5A). Within the hypothalamus, statistically significant Per2 mRNA rhythms were detected in vehicle-treated mice, while chemotherapy-treated mice exhibited rhythms in Arntl and Per2 gene expression (Fig. 5B). Furthermore, at circadian time (CT) 14, corresponding to the early subjective night, paclitaxel elevated Arntl (t₈ = 3.43, p < 0.01), Nr1d1 (t₈ = 2.30, p = 0.05), and Per2 hypothalamic gene expression (t₉ = 3.62, p < 0.01), and caused a trend towards elevated Per1 mRNA (p = 0.07, Fig. 5B). In the hippocampus, paclitaxel ablated Nr1d1 rhythms and increased Nr1d1 expression at CT 2 in the early subjective day (t₈ = 12.4, p < 0.0001, Fig. 5C). Hippocampal Per2 rhythms were not significantly altered by chemotherapy other than a modest increase in expression in the late subjective day (CT 10, p = 0.054, Fig. 5C). While statistically significant hippocampal Per1 rhythms were not observed in either vehicle- or chemotherapy-treated mice, chemotherapy increased Per1 in the late subjective night (CT 18, t₉ = 2.26, p = 0.05, Fig. 5C). In the frontal cortex, while all genes examined in both groups exhibited significant rhythmicity (p < 0.05), paclitaxel reduced the amplitude of Arntl, Nr1d1, and Per2 rhythms (Fig. 5D, Supplementary Table 2). These reduced rhythm amplitudes were due to significant changes at CT 2 (Arntl: t₉ = 2.52, p < 0.05; Nr1d1: t₉ = 2.87, p < 0.05; Per2: t₈ = 3.73, p < 0.01) and at CT18 for Per2 (U = 2, n₁ = 5, n₂ = 6, p < 0.05; Fig. 5D).

3.6. Chemotherapy ablates circulating corticosterone rhythms and attenuates rhythm amplitudes in adrenal gland clock gene expression in constant darkness.

Paclitaxel ablated 24-h rhythms in circulating CORT by elevating corticosterone concentrations at CT 2 (U = 0, n₁ = 5, n₂ = 6, p < 0.005), CT 6 (t₁₀ = 2.56, p < 0.05), and CT 18 (U = 2, n₁ = n₂ = 6, p < 0.01, Fig. 6A). However, no statistically significant rhythms were detected in any circulating cytokines or chemokines in constant darkness (Supplementary Fig. 2). Chemotherapy did not ablate adrenal gland clock gene rhythms, but paclitaxel decreased the amplitudes of Arntl, Nr1d1, Per1, and Star rhythms (Fig. 6B, Supplementary Table 2). Specifically, chemotherapy significantly altered adrenal gland clock gene expression at the following time points: Arntl at CT 2 (t₁₀ = 2.63, p < 0.05) and CT 14 (t₁₂ = 4.53, p < 0.001); Nr1d1 at CT 6 (t₉ = 3.18, p < 0.05) and CT 22 (U = 2, n₁ = 6, n₂ = 6, p < 0.01); Per1 at CT 2 (U = 0, n₁ = n₂ = 6, p < 0.005) and CT 18 (U = 5, n₁ = n₂ = 6, p < 0.05); Per2 at CT 2 (t₉ = 2.95, p < 0.05), CT 10 (t₁₀ = 2.39, p < 0.05), and CT 22 (U = 4, n₁ = n₂ = 6, p < 0.05); Star at CT 2 (t₁₀ = 3.21, p < 0.01) and CT 14 (t₉ = 2.30, p < 0.05, Fig. 6B).

3.7. Paclitaxel alters SCN and adrenal gland rhythms in vitro and induces lasting increases in pro-inflammatory cytokine expression.

After adjusting to a 12:12 LD light cycle, PER2::luciferase (PER2::LUC) mice were placed in constant darkness (DD) for two days before being injected with vehicle or paclitaxel (Fig. 7A). SCN slices cultured from mice treated with vehicle or paclitaxel in vivo did not exhibit different periods of PER2::LUC rhythms (p > 0.05), but slices from mice injected with vehicle but cultured with paclitaxel in the dish (vehicle + chemo in vitro) exhibited shorter PER2::LUC rhythms (Fig. 7B, F_2,10 = 4.68, p < 0.05; post-hoc comparisons between vehicle and vehicle + chemo in vitro: p < 0.05). Neither SCN phase nor amplitude were affected by paclitaxel treatment in vivo or in vitro (Fig. 7B, p > 0.05). In contrast, while the period (p = 0.07) and phase (p > 0.05) of adrenal PER2::LUC rhythms were not significantly different among groups, the amplitude of adrenal rhythms was significantly increased in paclitaxel-treated sections in vitro (Fig. 7C, F_2,8 = 4.46, p < 0.05).

In the culture media of SCN PER2::LUC slices, there were no statistically significant differences among groups in secreted IL-4 (p = 0.07) or IL-5 (p = 0.05), or IL-12p70 (Fig. 7D, p = 0.05). Moreover, SCN slices treated with paclitaxel in vitro did not increase the release of any cytokines or chemokines in culture media (Fig. 7D, post-hoc Bonferroni tests between vehicle and chemo in vitro p > 0.05). However, SCN slices from mice treated with paclitaxel in vivo increased culture media concentrations of IL-1β, IL-2, IL-6, IL-10, IL-12p70, IFN-γ, TNF-α, and CXCL1 (Fig. 7D, IL-1β: F_2,8, = 9.13, p < 0.01; IL-2: F_2,8, = 6.50, p < 0.05; IL-6: F_2,8, = 4.61, p < 0.05; IL-10: F_2,8, = 14.17, p < 0.005; IL-12p70: F_2,8, = 5.63, p < 0.05; IFN-γ: F_2,8, = 5.78, p < 0.05; TNF-α: F_2,8, = 22.25, p < 0.001; CXCL1: F_2,8, = 7.45, p < 0.05).

4. Discussion

The present study describes for the first time how paclitaxel, a common chemotherapeutic drug, disrupts circadian rhythms in behavior, physiology, and clock genes. Only two studies, to our knowledge, have assessed how cancer treatments may disrupt endogenous circadian circuitry (in constant darkness) in the SCN or voluntary wheel running rhythms. One study described flattened clock gene rhythms in the SCN by 5-fluorouracil chemotherapy (Terazono et al., 2008) and the other described the capacity of ɣ-radiation to induce a phase shift in free-running rhythms (Oklejewicz et al., 2008). The present study extends this work as a comprehensive investigation of paclitaxel effects on circadian rhythms in clock-controlled behavior, clock gene rhythms in multiple brain regions, altered PER2 protein oscillations in the SCN and adrenal glands, circulating corticosterone (CORT), and pro-inflammatory cytokine and chemokine secretion. Given the association among flattened cortisol slopes, circadian disruption to rest-activity cycles, and decreased long-term survival in cancer patients (Abercrombie et al., 2004; Adam et al., 2017; Mormont et al., 2000; Sephton et al., 2000; Zeitzer et al., 2016), the present results indicate that chemotherapy-induced inflammation likely contributes to the association between circadian disruption and decreased survival and increased comorbidities in cancer survivors.

We first examined the effects of paclitaxel on voluntary wheel running in a light-dark (LD) cycle to which circadian circuitry can be entrained. Paclitaxel decreased overall activity and the onset of activity during treatment, but not during recovery as evidenced by an interaction between day analyzed and drug. This transient response is likely due to the rapid metabolism of paclitaxel, as the phase angle of entrainment and wheel running activity returned to control levels within days of chemotherapy cessation. Notably, previous work demonstrated either a phase advance or phase delay depending on the time gemcitabine was administered in a LD cycle, which lasted for 2 d following treatment (Li and Levi, 2007). As we also did not see a long-lasting change in the total amount or onset of activity, it is possible that the entrainment provided by the light cycle prevented further disruption to locomotor rhythms caused by paclitaxel.

Because the strongest chemotherapy effects on activity onset occurred closest to the time of injections, we next looked at the effects of paclitaxel on rhythms in circadian gene expression in the brain in an LD light cycle. While the present study is the first to describe paclitaxel-induced clock gene changes in the brain, previous work has demonstrated the capacity for paclitaxel to induce inflammatory gene expression in the hypothalamus and hippocampus (Loman et al., 2019). Here, we observed a rhythm in hypothalamic Per1 from control mice in LD but not in chemotherapy-treated mice, and paclitaxel decreased the amplitude of hypothalamic Per2 rhythms. The peak phase of Per1/2 expression was consistent with previous work from whole hypothalamus in a LD cycle that also did not detect Arntl rhythms (Husse et al., 2017). Notably, rhythms were not observed in all clock genes under LD or DD conditions in mRNA obtained from whole hypothalamus. The SCN constituted a portion of this tissue, however, it is likely that somewhat asynchronous clock gene rhythms in other hypothalamic nuclei masked the detection of rhythms in clock genes from the whole hypothalamus. For instance, in a 12:12 LD cycle, the phase of peak Per1 expression is estimated near ZT 3-6 in the SCN (Albrecht et al., 1997; Chun et al., 2015; Jin et al., 1999), ZT 12 in the arcuate nucleus (Wang et al., 2017), ZT 12 in the paraventricular nucleus (Chun et al., 2015), and ZT 23 in the lateral hypothalamus (Opperhuizen et al., 2016). In this study, we dissected the whole hypothalamus in order to examine circadian gene rhythms in the brain region containing the master clock while simultaneously examining clock gene rhythms in the cortex, hippocampus, and adrenal glands, as well as corticosterone rhythms obtained from whole blood; thus, the present work captures the negative effects of paclitaxel on circadian transcriptional rhythmicity across a variety of tissues. However, given the capacity of paclitaxel to disrupt master clock-driven wheel running behavior and shorten PER2::LUC rhythms in the SCN (discussed below), it is likely that rhythms in clock genes would be detected if only the SCN were dissected from the hypothalamus.

In the hippocampus, Arntl rhythms were ablated by paclitaxel, whereas Nr1d1 and Per2 rhythms remained intact. The peak of Arntl and Per2 expression was comparable to previous studies of hippocampal clocks in a LD cycle (Chun et al., 2015; Jilg et al., 2010); however, these studies detected a significant hippocampal Per1 rhythm which was not robust in our LD experiment. In the frontal cortex, all clock genes were rhythmic in both treatment groups, but paclitaxel flattened the amplitudes of Nr1d1, Per1, and Per2 rhythms. The peak of cortical Arntl, Per1, and Per2 in this study was consistent with previous studies examining cortical clocks (Chun et al., 2015; Wakamatsu et al., 2001; Yang et al., 2007), and the time of peak expression was unaffected by paclitaxel. Notably, to our knowledge this is the first study describing Nr1d1 hippocampal rhythms in a LD cycle, which peaked between the mid-day to day-night transition period in both control and chemotherapy-treated mice. Taken together, these changes indicate that brain rhythms were not ablated by paclitaxel, but that the amplitudes of several genes were decreased by chemotherapy treatment in a LD cycle. Given the importance of hippocampal and cortical clocks to cognitive function (Bering et al., 2017; Snider et al., 2016), it is possible that these decreased clock gene amplitudes may correlate with paclitaxel-induced cognitive deficits (Loman et al., 2019) by making the clock gene rhythmicity less pronounced in these brain regions.

We next examined the adrenal glands as flattened cortisol rhythms have been commonly reported in cancer patients (Abercrombie et al., 2004; Sephton et al., 2000). In a LD cycle, strong circadian rhythms were observed in all circadian genes examined in the adrenal glands, indicating that the adrenal clock was not rendered arrhythmic by paclitaxel. However, paclitaxel flattened the amplitudes of Nr1d1, Per2, and Star rhythms. Intriguingly, while paclitaxel reduced the amplitude of these clock gene rhythms, it increased the amplitude of circulating CORT rhythms. It is therefore possible that paclitaxel increased circulating CORT in a LD cycle by affecting clocks in the pituitary gland to mediate adrenocorticotrophic hormone (ACTH) release upstream, but this was not examined in the present study. Notably, in Per2/Cry1 clock gene mutant mice that lack a functional circadian clock, both ACTH and CORT rhythms are flattened in LD and DD (Oster et al., 2006b). Further work is merited to understand the relationship between the paclitaxel-induced changes to the amplitudes of adrenal Star expression and circulating corticosterone in a LD cycle.

Next, voluntary wheel running rhythms in constant darkness (DD) were examined to determine the extent to which paclitaxel affected free-running rhythms in the absence of entrainment signals. By eliminating entrainment signals, the masking effects of light on the SCN-driven behavior are removed, as even mice with global Bmal1 deletion that lack a functional clock maintain rhythmic behavior in an LD cycle (Bunger et al., 2000). While paclitaxel decreased total activity during treatment and the week following treatment, as has been observed in studies of paclitaxel and other chemotherapy drugs studied during LD (Li and Levi, 2007; Li et al., 2002; Ray et al., 2011), the lengthening of this master clock-driven behavior occurred during the recovery weeks following chemotherapy treatment. This indicates a potential residual effect of paclitaxel on locomotor rhythms that was not observed during an LD cycle. Notably, while other groups have reported effects of γ-radiation inducing phase shifts in DD (Oklejewicz et al., 2008) and chemotherapy effects on LD wheel running activity (Li and Levi, 2007; Li et al., 2002), to our knowledge this is the first evidence of chemotherapy lengthening the free-running rhythm of SCN-regulated wheel running in constant darkness. Moreover, fatigue was more pronounced during the recovery period in paclitaxel-treated mice in a DD cycle compared to mice in a LD cycle. Given the more profound effects of paclitaxel on wheel running rhythms and fatigue in constant darkness compared to a light-dark cycle, this may indicate that strong entrainment cues (e.g., morning light therapy) could attenuate chemotherapy-induced circadian dysrhythmia and fatigue.

In order to assess potential paclitaxel-induced changes in circadian gene expression underlying the observed disruptions in wheel running, brain tissue was collected around the clock after the chemotherapy paradigm. In constant darkness, Per2 was the only hypothalamic gene to be significantly rhythmic, which was the case for both control and chemotherapy-treated groups. Contrary to our hypothesis, hypothalamic Arntl expression was only significantly rhythmic in paclitaxel-treated mice. Notably, the peak expression of Arntl and Per2 in this study was consistent with previous work analyzing whole hypothalamus in constant darkness (Zhang et al., 2014). However, paclitaxel increased hypothalamic Arntl, Nr1d1, and Per2 expression in the early subjective night (CT 14) compared to vehicle, indicating that chemotherapy modestly, but consistently, altered circadian gene expression in the whole hypothalamus. As paclitaxel did not affect clock gene expression in the early night in mice in a light-dark cycle, it is possible that paclitaxel causes more dramatic effects on hypothalamic clock gene rhythmicity in the absence of entrainment signals.

Next, clock gene expression in the hippocampus and cortex were examined due to their role in cognition, as cognitive deficits are commonly reported in cancer patients during chemotherapy treatment (reviewed in Ahles et al., 2012). In the hippocampus, paclitaxel abolished Nr1d1 rhythms and increased Per1 expression at CT 18. The peak phase of hippocampal Nr1d1 and Per2 in this study was also consistent with previous work in constant darkness (Bering et al., 2017; Ma et al., 2016), though notably these studies had placed animals into constant darkness for two days compared to the 19 days in this study. In contrast to the hippocampus, significant Arntl, Nr1d, Per1, and Per2 rhythms were detected in the frontal cortex, with phases of peak expression near previously reported times of highest expression (Bering et al., 2017; Wakamatsu et al., 2001). However, paclitaxel reduced the amplitudes of cortical Arntl, Nr1d1, and Per2, similarly to what was observed in LD conditions. Given the paclitaxel-induced changes to hippocampal and cortical rhythms in both LD and DD, further work is merited to understand which other clock-controlled genes in the hippocampus and cortex are altered by chemotherapy, as these genes may be important for cognitive, mood, or fatigue comorbidities described in cancer patients.

In the present study, the effects of paclitaxel on adrenal clocks and CORT release was assessed in constant darkness. While previous work demonstrated irinotecan chemotherapy-induced decreases in CORT release in female mice (but increases in male mice) in an LD cycle (Ahowesso et al., 2011), here we expand on these findings by demonstrating that paclitaxel elevated CORT across a 24-h cycle in DD in female mice, and thus ablated its circadian rhythm. This chemotherapy-induced loss of CORT rhythmicity parallels, in a tumor-free mouse, changes to diurnal CORT slope observed in cancer patients (Abercrombie et al., 2004; Cuneo et al., 2017; Schmidt et al., 2016; Sephton et al., 2000; Touitou et al., 1996). In conjunction with a loss of rhythmic circulating CORT, Arntl, Nr1d1, Per1/1, and Star rhythm amplitudes were all decreased in the adrenals by paclitaxel while the phases of peak expression of these genes were consistent with previous reports (Oster et al., 2006a; Son et al., 2008). Given that circulating CORT rhythms were not impaired in a LD cycle, it is possible that stronger entrainment cues (i.e., light) may override paclitaxel-induced ablation of CORT rhythms. These preclinical results may suggest that circadian therapies (e.g., bright light therapy) could be used as an intervention to combat changes to cortisol rhythmicity in humans, as the loss of diurnal CORT rhythms correlates with poorer long-term cancer survival (Sephton et al., 2000). Furthermore, dysregulated CORT release may also contribute to changes in hippocampal and cortical clocks, as time-of-day differences in glucocorticoid receptor (GR) and mineralocorticoid receptor (MR) binding to the Per1 promoter in hippocampal tissue have been reported (Mifsud and Reul, 2016). However, it is yet to be determined whether chemotherapy directly alters GR/MR binding to Per1, and the extent to which paclitaxel-induced changes in hippocampal and cortical clocks modulate the functions of those brain regions.

Finally, to determine whether paclitaxel was directly or indirectly affecting SCN and adrenal gland rhythms and pro-inflammatory cytokine and chemokine release, PER2::LUC cultures were analyzed from mice previously treated with paclitaxel in vivo and compared to vehicle-treated mice with paclitaxel present in vitro. Intriguingly, paclitaxel in vitro decreased the period of SCN rhythmicity while trending towards a lengthened adrenal gland rhythm and increased the amplitude of adrenal gland rhythms while leaving the amplitude of SCN rhythms unaltered. These disparate effects to SCN and adrenal gland rhythms suggest that paclitaxel has the capacity to directly alter circadian rhythmicity in both tissues which may have functional consequences in cancer patient populations. Moreover, mice treated with paclitaxel in vivo induced SCN release of a number of pro-inflammatory cytokines previously associated with cancer treatment symptoms, including IL-1β and TNF-α (Smith et al., 2014). As this effect was not sustained in SCN slices treated with paclitaxel acutely, this indicated that paclitaxel induced inflammation in the SCN that persisted postmortem. As TNF-α can interfere with E-box-mediated clock gene transcription (Cavadini et al., 2007), it is possible that paclitaxel-induced peripheral and central inflammation underlies the circadian disruption described in this study (Loman et al., 2019). Furthermore, as slices treated with paclitaxel in vitro also decreased SCN period length while cytokine and chemokine concentrations remained unchanged, this suggests additional mechanisms underlie paclitaxel-induced alterations to circadian rhythmicity. Future work is needed to understand which cytokine and chemokine signals may underlie paclitaxel-induced changes to circadian rhythmicity, which may provide additional drug targets to prevent the observed circadian alterations.

Together, the present study demonstrates that chemotherapy disrupts circadian rhythms in locomotion and clock gene expression independently of tumor biology. While previous studies have highlighted that the efficacy of chemotherapy as an anti-cancer treatment is dependent on the time it is administered (reviewed in Smith et al., 2014), here we demonstrate that chemotherapy conversely alters existing circadian rhythms throughout the body. Given the capacity for non-CNS tumors to also alter CORT rhythms, wheel running rhythms, and hypothalamic circadian gene expression (Sullivan et al., 2019), the investigation of combinatorial effects of tumors and paclitaxel on circadian rhythms in mice is warranted. In understanding the mechanisms of circadian disruption in preclinical models, this work may lead to more tailored, circadian-based treatment strategies for improving cancer patient quality of life and long-term survival.

Highlights.

• Paclitaxel lengthens wheel running rhythms in constant darkness

• Paclitaxel alters clock gene rhythmicity in the SCN and multiple brain regions

• Paclitaxel elevates circulating corticosterone and ablates its rhythmicity

• Paclitaxel dysregulates the amplitude of circadian genes in the adrenal glands

• Paclitaxel causes lasting SCN pro-inflammatory cytokine and chemokine secretion